arXiv subjects

David Zhang

Publications and source records attributed to David Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Prompt-Driven Exploration

Exploration is essential to RL since a policy cannot improve by repeatedly sampling the behaviors it already prefers. Standard methods inject stochasticity in the action space, but such jitter only yields rollouts close to the original. Escaping a weak policy often requires global perturbations that action noise cannot produce. Large language models (LLMs) and vision-language-action (VLA) models offer a pathway: they condition the policy on a natural language prompt, and since the rollout follows from it, modifying the prompt induces global changes. The challenge is finding prompts that induce useful global changes. With a weak policy that rarely succeeds, reward is too sparse to select on. Our idea is to refine prompts from the rollouts themselves: a vision-language model (VLM) reasons over the rollout video, diagnoses how the policy responded, and rewrites the prompt to elicit better behavior next time. This procedure resembles posterior sampling, a classical RL exploration framework, at the level of prompts: the VLM maintains an implicit distribution over useful prompts and updates it from observed rollouts. We call this strategy Prompt-Driven Exploration (PDE). Across manipulation and reasoning tasks, PDE enables RL to learn successful policies even from zero-reward starts, and improves sample efficiency more broadly. Our website is available at https://xinyunsunshine.github.io/prompt-rl.

cs.LG↗

arXiv2026

A Deep Dive into Scaling RL for Code Generation with Synthetic Data and Curricula

Reinforcement learning (RL) has emerged as a powerful paradigm for improving large language models beyond supervised fine-tuning, yet sustaining performance gains at scale remains an open challenge, as data diversity and structure, rather than volume alone, become the limiting factor. We address this by introducing a scalable multi-turn synthetic data generation pipeline in which a teacher model iteratively refines problems based on in-context student performance summaries, producing structured difficulty progressions without any teacher fine-tuning. Compared to single-turn generation, this multi-turn approach substantially improves the yield of valid synthetic problems and naturally produces stepping stones, i.e. easier and harder variants of the same core task, that support curriculum-based training. We systematically study how task difficulty, curriculum scheduling, and environment diversity interact during RL training across the Llama3.1-8B Instruct and Qwen3-8B Base model families, with additional scaling experiments on Qwen2.5-32B. Our results show that synthetic augmentation consistently improves in-domain code and in most cases out-of-domain math performance, and we provide empirical insights into how curriculum design and data diversity jointly shape RL training dynamics.

cs.LG↗

arXiv2026

Prompt Tuning for CLIP on the Pretrained Manifold

Prompt tuning introduces learnable prompt vectors that adapt pretrained vision-language models to downstream tasks in a parameter-efficient manner. However, under limited supervision, prompt tuning alters pretrained representations and drives downstream features away from the pretrained manifold toward directions that are unfavorable for transfer. This drift degrades generalization. To address this limitation, we propose ManiPT, a framework that performs prompt tuning on the pretrained manifold. ManiPT introduces cosine consistency constraints in both the text and image modalities to confine the learned representations within the pretrained geometric neighborhood. Furthermore, we introduce a structural bias that enforces incremental corrections, guiding the adaptation along transferable directions to mitigate reliance on shortcut learning. From a theoretical perspective, ManiPT alleviates overfitting tendencies under limited data. Our experiments cover four downstream settings: unseen-class generalization, few-shot classification, cross-dataset transfer, and domain generalization. Across these settings, ManiPT achieves higher average performance than baseline methods. Notably, ManiPT provides an explicit perspective on how prompt tuning overfits under limited supervision.

cs.CV↗

arXiv2026

Dynamic Prior Thompson Sampling for Cold-Start Exploration in Recommender Systems

Cold-start exploration is a core challenge in large-scale recommender systems: new or data-sparse items must receive traffic to estimate value, but over-exploration harms users and wastes impressions. In practice, Thompson Sampling (TS) is often initialized with a uniform Beta(1,1) prior, implicitly assuming a 50% success rate for unseen items. When true base rates are far lower, this optimistic prior systematically over-allocates to weak items. The impact is amplified by batched policy updates and pipeline latency: for hours, newly launched items can remain effectively "no data," so the prior dominates allocation before feedback is incorporated. We propose Dynamic Prior Thompson Sampling, a prior design that directly controls the probability that a new arm outcompetes the incumbent winner. Our key contribution is a closed-form quadratic solution for the prior mean that enforces P(X_j > Y_k) = epsilon at introduction time, making exploration intensity predictable and tunable while preserving TS Bayesian updates. Across Monte Carlo validation, offline batched simulations, and a large-scale online experiment on a thumbnail personalization system serving millions of users, dynamic priors deliver precise exploration control and improved efficiency versus a uniform-prior baseline.

cs.LG↗

arXiv2026

A Cosine Network for Image Super-Resolution

Deep convolutional neural networks can use hierarchical information to progressively extract structural information to recover high-quality images. However, preserving the effectiveness of the obtained structural information is important in image super-resolution. In this paper, we propose a cosine network for image super-resolution (CSRNet) by improving a network architecture and optimizing the training strategy. To extract complementary homologous structural information, odd and even heterogeneous blocks are designed to enlarge the architectural differences and improve the performance of image super-resolution. Combining linear and non-linear structural information can overcome the drawback of homologous information and enhance the robustness of the obtained structural information in image super-resolution. Taking into account the local minimum of gradient descent, a cosine annealing mechanism is used to optimize the training procedure by performing warm restarts and adjusting the learning rate. Experimental results illustrate that the proposed CSRNet is competitive with state-of-the-art methods in image super-resolution.

cs.CV↗

arXiv2025

Toward Training Superintelligent Software Agents through Self-Play SWE-RL

While current software agents powered by large language models (LLMs) and agentic reinforcement learning (RL) can boost programmer productivity, their training data (e.g., GitHub issues and pull requests) and environments (e.g., pass-to-pass and fail-to-pass tests) heavily depend on human knowledge or curation, posing a fundamental barrier to superintelligence. In this paper, we present Self-play SWE-RL (SSR), a first step toward training paradigms for superintelligent software agents. Our approach takes minimal data assumptions, only requiring access to sandboxed repositories with source code and installed dependencies, with no need for human-labeled issues or tests. Grounded in these real-world codebases, a single LLM agent is trained via reinforcement learning in a self-play setting to iteratively inject and repair software bugs of increasing complexity, with each bug formally specified by a test patch rather than a natural language issue description. On the SWE-bench Verified and SWE-Bench Pro benchmarks, SSR achieves notable self-improvement (+10.4 and +7.8 points, respectively) and consistently outperforms the human-data baseline over the entire training trajectory, despite being evaluated on natural language issues absent from self-play. Our results, albeit early, suggest a path where agents autonomously gather extensive learning experiences from real-world software repositories, ultimately enabling superintelligent systems that exceed human capabilities in understanding how systems are constructed, solving novel challenges, and autonomously creating new software from scratch.

cs.SE↗

arXiv2025

Jailbreak-Zero: A Path to Pareto Optimal Red Teaming for Large Language Models

This paper introduces Jailbreak-Zero, a novel red teaming methodology that shifts the paradigm of Large Language Model (LLM) safety evaluation from a constrained example-based approach to a more expansive and effective policy-based framework. By leveraging an attack LLM to generate a high volume of diverse adversarial prompts and then fine-tuning this attack model with a preference dataset, Jailbreak-Zero achieves Pareto optimality across the crucial objectives of policy coverage, attack strategy diversity, and prompt fidelity to real user inputs. The empirical evidence demonstrates the superiority of this method, showcasing significantly higher attack success rates against both open-source and proprietary models like GPT-40 and Claude 3.5 when compared to existing state-of-the-art techniques. Crucially, Jailbreak-Zero accomplishes this while producing human-readable and effective adversarial prompts with minimal need for human intervention, thereby presenting a more scalable and comprehensive solution for identifying and mitigating the safety vulnerabilities of LLMs.

cs.CL↗

arXiv2025

A Perception CNN for Facial Expression Recognition

Convolutional neural networks (CNNs) can automatically learn data patterns to express face images for facial expression recognition (FER). However, they may ignore effect of facial segmentation of FER. In this paper, we propose a perception CNN for FER as well as PCNN. Firstly, PCNN can use five parallel networks to simultaneously learn local facial features based on eyes, cheeks and mouth to realize the sensitive capture of the subtle changes in FER. Secondly, we utilize a multi-domain interaction mechanism to register and fuse between local sense organ features and global facial structural features to better express face images for FER. Finally, we design a two-phase loss function to restrict accuracy of obtained sense information and reconstructed face images to guarantee performance of obtained PCNN in FER. Experimental results show that our PCNN achieves superior results on several lab and real-world FER benchmarks: CK+, JAFFE, FER2013, FERPlus, RAF-DB and Occlusion and Pose Variant Dataset. Its code is available at https://github.com/hellloxiaotian/PCNN.

cs.CV↗

arXiv2025

Unsupervised Robust Domain Adaptation: Paradigm, Theory and Algorithm

Unsupervised domain adaptation (UDA) aims to transfer knowledge from a label-rich source domain to an unlabeled target domain by addressing domain shifts. Most UDA approaches emphasize transfer ability, but often overlook robustness against adversarial attacks. Although vanilla adversarial training (VAT) improves the robustness of deep neural networks, it has little effect on UDA. This paper focuses on answering three key questions: 1) Why does VAT, known for its defensive effectiveness, fail in the UDA paradigm? 2) What is the generalization bound theory under attacks and how does it evolve from classical UDA theory? 3) How can we implement a robustification training procedure without complex modifications? Specifically, we explore and reveal the inherent entanglement challenge in general UDA+VAT paradigm, and propose an unsupervised robust domain adaptation (URDA) paradigm. We further derive the generalization bound theory of the URDA paradigm so that it can resist adversarial noise and domain shift. To the best of our knowledge, this is the first time to establish the URDA paradigm and theory. We further introduce a simple, novel yet effective URDA algorithm called Disentangled Adversarial Robustness Training (DART), a two-step training procedure that ensures both transferability and robustness. DART first pre-trains an arbitrary UDA model, and then applies an instantaneous robustification post-training step via disentangled distillation.Experiments on four benchmark datasets with/without attacks show that DART effectively enhances robustness while maintaining domain adaptability, and validate the URDA paradigm and theory.

cs.LG↗

arXiv2025

Deterministically Simulating Barely Random Algorithms in the Random-Order Arrival Model

Interest in the random-order model (ROM) leads us to initiate a study of utilizing random-order arrivals to extract random bits with the goal of derandomizing algorithms. Besides producing simple algorithms, simulating random bits through random arrivals enhances our understanding of the comparative strength of randomized online algorithms (with adversarial input sequences) and deterministic algorithms in the ROM. We consider three $1$-bit randomness extraction processes. Our best extraction process returns a bit with a worst-case bias of $2 - \sqrt{2} \approx 0.585$ and operates under the mild assumption that there exist at least two distinct items in the input. We motivate the applicability of this process by using it to simulate a number of barely random algorithms for weighted interval selection (single-length with arbitrary weights, as well as monotone, C-benevolent and D-benevolent weighted instances), the proportional and general knapsack problems, job throughput scheduling, and makespan minimization. It is well known that there are many applications where a deterministic ROM algorithm significantly outperforms any randomized online algorithm (in terms of competitive ratios). The classic example is that of the secretary problem. We ask the following fundamental question: Is there any application for which a randomized algorithm outperforms any deterministic ROM algorithm? Motivated by this question, we view our randomness extraction applications as a constructive approach toward understanding the relationship between randomized online algorithms and deterministic algorithms in the ROM.

cs.DS↗

arXiv2025

GTAlign: Game-Theoretic Alignment of LLM Assistants for Social Welfare

Large Language Models (LLMs) have achieved remarkable progress in reasoning, yet sometimes produce responses that are suboptimal for users in tasks such as writing, information seeking, or providing practical guidance. Conventional alignment practices typically assume that maximizing model reward also maximizes user welfare, but this assumption frequently fails in practice: models may over-clarify or generate overly verbose reasoning when users prefer concise answers. Such behaviors resemble the prisoner's dilemma, where individually rational choices lead to socially suboptimal outcomes. The fundamental challenge is the lack of a principled decision making mechanism that mutually benefits both the LLM and the user. We propose Game-Theoretic Alignment (GTAlign), an alignment framework that integrates game-theoretic decision making into both reasoning and training. During reasoning, the model explicitly treats user-LLM interaction as a strategic game: it constructs payoff matrices within its reasoning chain to estimate welfare for both itself and the user, and then selects actions that are mutually beneficial. During training, we introduce a social welfare reward that reinforces cooperative responses, aligning model behavior with socially efficient outcomes. In addition, we introduce an inference technique that leverages game-theoretic reasoning to dynamically adapt LLM's response when pricing policies of LLM service change. Extensive experiments demonstrate that GTAlign substantially improves reasoning efficiency, answer quality, and social welfare compared to baselines across diverse tasks. The code is available at https://github.com/ulab-uiuc/GTAlign .

cs.AI↗

arXiv2025

CWM: An Open-Weights LLM for Research on Code Generation with World Models

We release Code World Model (CWM), a 32-billion-parameter open-weights LLM, to advance research on code generation with world models. To improve code understanding beyond what can be learned from training on static code alone, we mid-train CWM on a large amount of observation-action trajectories from Python interpreter and agentic Docker environments, and perform extensive multi-task reasoning RL in verifiable coding, math, and multi-turn software engineering environments. With CWM, we provide a strong testbed for researchers to explore the opportunities world modeling affords for improving code generation with reasoning and planning in computational environments. We present first steps of how world models can benefit agentic coding, enable step-by-step simulation of Python code execution, and show early results of how reasoning can benefit from the latter. CWM is a dense, decoder-only LLM trained with a context size of up to 131k tokens. Independent of its world modeling capabilities, CWM offers strong performance on general coding and math tasks: it reaches pass@1 scores of 65.8% on SWE-bench Verified (with test-time scaling), 68.6% on LiveCodeBench, 96.6% on Math-500, and 76.0% on AIME 2024. To support further research on code world modeling, we release model checkpoints after mid-training, SFT, and RL.

cs.SE↗

arXiv2025

The Challenge of Teaching Reasoning to LLMs Without RL or Distillation

Reasoning-capable language models achieve state-of-the-art performance in diverse complex tasks by generating long, explicit Chain-of-Thought (CoT) traces. While recent works show that base models can acquire such reasoning traces via reinforcement learning or distillation from stronger models like DeepSeek-R1, previous works demonstrate that even short CoT prompting without fine-tuning is able to improve reasoning. We ask whether long CoT can be induced in a base model using only prompting or minimal tuning. Using just 20 long CoT examples from the reasoning model \texttt{QwQ-32B-Preview}, we lightly fine-tune the base model \texttt{Qwen2.5-32B}. The resulting model outperforms the much larger \texttt{Qwen2.5-Math-72B-Instruct}, showing that a handful of high-quality examples can unlock strong reasoning capabilities. We further explore using CoT data from non-reasoning models and human annotators, enhanced with prompt engineering, multi-pass editing, and structural guidance. However, neither matches the performance of reasoning model traces, suggesting that certain latent qualities of expert CoT are difficult to replicate. We analyze key properties of reasoning data, such as problem difficulty, diversity, and answer length, that influence reasoning distillation. While challenges remain, we are optimistic that carefully curated human-written CoT, even in small quantities, can activate reasoning behaviors in base models. We release our human-authored dataset across refinement stages and invite further investigation into what makes small-scale reasoning supervision so effective.

cs.AI↗

arXiv2025

A Tree-guided CNN for image super-resolution

Deep convolutional neural networks can extract more accurate structural information via deep architectures to obtain good performance in image super-resolution. However, it is not easy to find effect of important layers in a single network architecture to decrease performance of super-resolution. In this paper, we design a tree-guided CNN for image super-resolution (TSRNet). It uses a tree architecture to guide a deep network to enhance effect of key nodes to amplify the relation of hierarchical information for improving the ability of recovering images. To prevent insufficiency of the obtained structural information, cosine transform techniques in the TSRNet are used to extract cross-domain information to improve the performance of image super-resolution. Adaptive Nesterov momentum optimizer (Adan) is applied to optimize parameters to boost effectiveness of training a super-resolution model. Extended experiments can verify superiority of the proposed TSRNet for restoring high-quality images. Its code can be obtained at https://github.com/hellloxiaotian/TSRNet.

eess.IV↗

arXiv2025

A Dynamic Transformer Network for Vehicle Detection

Stable consumer electronic systems can assist traffic better. Good traffic consumer electronic systems require collaborative work between traffic algorithms and hardware. However, performance of popular traffic algorithms containing vehicle detection methods based on deep networks via learning data relation rather than learning differences in different lighting and occlusions is limited. In this paper, we present a dynamic Transformer network for vehicle detection (DTNet). DTNet utilizes a dynamic convolution to guide a deep network to dynamically generate weights to enhance adaptability of an obtained detector. Taking into relations of different information account, a mixed attention mechanism based channel attention and Transformer is exploited to strengthen relations of channels and pixels to extract more salient information for vehicle detection. To overcome the drawback of difference in an image account, a translation-variant convolution relies on spatial location information to refine obtained structural information for vehicle detection. Experimental results illustrate that our DTNet is competitive for vehicle detection. Code of the proposed DTNet can be obtained at https://github.com/hellloxiaotian/DTNet.

cs.CV↗

arXiv2025

Domain Gating Ensemble Networks for AI-Generated Text Detection

As state-of-the-art language models continue to improve, the need for robust detection of machine-generated text becomes increasingly critical. However, current state-of-the-art machine text detectors struggle to adapt to new unseen domains and generative models. In this paper we present DoGEN (Domain Gating Ensemble Networks), a technique that allows detectors to adapt to unseen domains by ensembling a set of domain expert detector models using weights from a domain classifier. We test DoGEN on a wide variety of domains from leading benchmarks and find that it achieves state-of-the-art performance on in-domain detection while outperforming models twice its size on out-of-domain detection. We release our code and trained models to assist in future research in domain-adaptive AI detection.

cs.CL↗

arXiv2025

Compact Recurrent Transformer with Persistent Memory

The Transformer architecture has shown significant success in many language processing and visual tasks. However, the method faces challenges in efficiently scaling to long sequences because the self-attention computation is quadratic with respect to the input length. To overcome this limitation, several approaches scale to longer sequences by breaking long sequences into a series of segments, restricting self-attention to local dependencies between tokens within each segment and using a memory mechanism to manage information flow between segments. However, these approached generally introduce additional compute overhead that restricts them from being used for applications where limited compute memory and power are of great concern (such as edge computing). We propose a novel and efficient Compact Recurrent Transformer (CRT), which combines shallow Transformer models that process short local segments with recurrent neural networks to compress and manage a single persistent memory vector that summarizes long-range global information between segments. We evaluate CRT on WordPTB and WikiText-103 for next-token-prediction tasks, as well as on the Toyota Smarthome video dataset for classification. CRT achieves comparable or superior prediction results to full-length Transformers in the language datasets while using significantly shorter segments (half or quarter size) and substantially reduced FLOPs. Our approach also demonstrates state-of-the-art performance on the Toyota Smarthome video dataset.

cs.LG↗

arXiv2025

Non-Markovian Discrete Diffusion with Causal Language Models

Discrete diffusion models offer a flexible, controllable approach to structured sequence generation, yet they still lag behind causal language models in expressive power. A key limitation lies in their reliance on the Markovian assumption, which restricts each step to condition only on the current state, leading to potential uncorrectable error accumulation. In this paper, we introduce CaDDi (Causal Discrete Diffusion Model), a discrete diffusion model that conditions on the entire generative trajectory, thereby lifting the Markov constraint and allowing the model to revisit and improve past states. By unifying sequential (causal) and temporal (diffusion) reasoning in a single non-Markovian transformer, CaDDi also treats standard causal language models as a special case and permits the direct reuse of pretrained LLM weights with no architectural changes. Empirically, CaDDi outperforms state-of-the-art discrete diffusion baselines on natural-language benchmarks, substantially narrowing the remaining gap to large autoregressive transformers.

cs.LG↗